Keywords

1 Introduction

As companies adjust to the post-pandemic work-life, managers have been grappling with whether and how to bring employees back to the office, and many companies offer a hybrid work solution. Many employees see work location flexibility as a bonus on par with increased salary [3]. However, many experience difficulties related to communication, collaboration, and cooperation with other team members [8, 17] found lack of communication, collaboration and coordination to be the cause of social debt. Lack of coordination is a common organizational challenge and has been found to be a cause of process debt [15].

Agile development at scale introduces new challenges. For example, there are more uncertainty, complexity, and dependencies between projects and teams, and thus coordination especially becomes a challenge [6]. Furthermore, the high degree of complexity and dependencies across teams threatens team autonomy [12]. A recent longitudinal study exploring coordination mechanisms in large-scale agile revealed that these mechanisms evolve in response to external and internal change events and that implementing the appropriate coordination mechanisms can significantly enhance the organization’s resilience [4]. We hereafter use the umbrella term organizational debt to include both social and process debt. We aimed to understand how challenges with coordination, that cause both process and social debt, can be managed by the use of coordination mechanisms. We explored the following research question:

RQ1: “What coordination mechanisms are used to manage organizational debt challenges in large-scale agile?”

2 Methodology

The research was carried out in the case organisation “PubTrans”, which is an organisation responsible for the software development project of a platform for public transportation in Norway. The project has existed since 2016, and the first author was hired as an IT consultant in the project from the end of 2019.

The company could be defined as large-scale, as it has seventeen development teams ranging between five and eighteen team members, each with their own responsibility area, and together working toward develo** the same products. The teams are able to choose their own tools, technology, agile methods and processes, and could thus be considered autonomous. As the pandemic restrictions were lifted in the middle of 2021, the company decided to experiment with a hybrid working arrangement, as many of its employees enjoyed the flexibility of being able to work remotely. Teams at PubTrans were allowed to design their own approach to the hybrid working arrangement, due to a high degree of autonomy. As a result, the teams continuously experimented with new coordination strategies throughout the pandemic and post-pandemic period.

We carried out and transcribed 10 semi-structured interviews with developers and designers, spanning 7 different development teams at PubTrans. Also, we collected artefacts which were relevant to the hybrid working arrangement, such as Slack logs and documentation from Jira, Confluence, Miro and Microsoft Teams. This was done in order to enable better data triangulation, and to improve the validity and reliability of the case study. The research started in January 2022, and an overview of the research timeline is illustrated in Fig. 1.

Fig. 1.
figure 1

Timeline of research

Full size image

When chosing informants, we minimized sampling bias through diversity and variation as selection criteria. We initially compiled a list of all potential informants within PubTrans, encompassing about 100 individuals engaged in software development. Given diverse experiences across roles - developers, designers, team leads, and managers - a comprehensive exploration of all roles was impractical for this study’s scope. Consequently, we focused on developers and designers at PubTrans, offering a diverse skillset within the informant group while maintaining manageable interview scope. In the preliminary research phase, it became evident that autonomous teams exhibited varying approaches to the hybrid work arrangement. To capture a spectrum of perspectives, we opted to select one or two representatives from several teams, facilitating the gathering of diverse narratives. Additionally, we deliberately chose informants with varying experience levels, personalities, life situations, and work settings (home or office).

A thematic coding analysis approach was taken when analyzing the data. We first created root nodes on NVIVO following the theoretical framework, and deductively generated the initial codes. However, we also inductively generated codes in order to stay close to the data. This resulted in a list of benefits and challenges caused by the hybrid working arrangement, and a list of coordination mechanisms identified based on the model by [22].

From January 2022 to March 2022, the first author assumed the role of a participant observer within the PubTrans organization. The participation included attending planned and impromptu meetings, formal discussions, informal interactions, and team collaborative efforts. Additionally, engagement extended to social gatherings like Friday gatherings and seminars. These activities provided a comprehensive understanding of the intricacies inherent to the PubTrans company. This understanding was particularly profound during the period of hybrid work arrangements. Importantly, this active involvement facilitated the establishment of rapport with key individuals before the subsequent interview phase.

3 Results

3.1 Shared Mental Models

We found one major challenge to be maintaining a shared mental model of the location and availability of other team members within the team. First, many informants were unsure about how many of their coworkers will be at the office on any given workday. Most did not see a reason to travel to the PubTrans office when their collaborators were working remotely, and only wished to co-locate if enough team members were at the office. Some felt lonely when staying at the office without other team members, as they did not know many other employees. Many wished to adjust their co-location plans according to the other team members, but this was difficult as most team members did not document their co-location plans. The majority decided their work location right before the work day started. “Yesterday I went to work, and there was no one there, and that wasn’t so fun. I could’ve just as well stayed at home” (Interview C2). Likewise, it was even more difficult for the employees to find the co-location plans of other teams. Most employees informed about their co-location plans in private team Slack channels, which were hidden from those outside the team.

Secondly, the informants experienced difficulties accommodating the other work mode when working from separate locations, due to a lack of shared awareness. This created challenges related to inclusivity. For example, those who were co-located at the office could often forget to accommodate to those working remotely, and those alone at the office could experience difficulties trying to participate in all the digital activities in an open office landscape. As a result, information shared between the co-located team members may not reach those working remotely. Similar challenges were also reported from hybrid meetings. Those who were co-located sometimes had informal conversations which excluded the digital participants. These conversations could be disruptive to the remote participants, and the remote participants also missed out on important information and decisions.

3.2 Team Coordination

We identified team coordination as a major organizational debt challenge at PubTrans, with hybrid work increasing meeting complexity and communication barriers. The interviewees reported fewer ad-hoc meetings in hybrid teams due to reduced co-location time. Initiating video calls without prior planning on Slack was uncommon, unlike in-person interactions at the office. Ad-hoc meetings mostly occurred when teams were co-located. “It’s a bit more painful when you’re at home. You always have to call. It’s such a big step [...] it feels like you’re interrupting others way more. Like, ‘oh my god, now there’s another message or a direct phone call from him’. It feels so dramatic” (Interview B1).

Secondly, the frequency of communication between teams was significantly reduced. Prior to the pandemic, teams working on similar domains at PubTrans were placed in the vicinity of each other, thus promoting the collaboration between these two teams. Similarly, task forces (i.e. a temporary team consisting of members from different teams working on the same feature) were built prior and during the pandemic as an inter-team coordination mechanism. This had however disappeared as a result of the hybrid work arrangement, as the teams no longer came to the office on the same days.

The employees also experienced longer feedback loops when collaborating on the same task, compared to before the pandemic. This was due to the increased barriers to initiating ad-hoc communication. In hybrid teams, conversations had to be more explicitly planned and executed using communication tools such as Slack and Microsoft Teams. Setting up the tools and waiting for asynchronous answers resulted in longer waiting time, and longer feedback loops during the collaboration sessions. This increased the threshold for asking questions, and decreased the coordination efficiency between team members.

3.3 Team Cohesion

Team cohesion was also identified as a major challenge caused by hybrid work. The team members reported to have decreased levels of attachment to the team, due to a lack of face-to-face social activities and informal conversations. It was more difficult to carry out informal conversations with the entire team in hybrid teams, as a result of the reduced and mismatched co-location time. Communication via tools such as Slack often felt impersonal, as they lacked additional dimensions such as body language. Furthermore, prior to the pandemic, the employees often gathered for dinners and other social activities after work. The frequency of social activities had drastically decreased after the pandemic.

Finally, despite encouragements to co-locate on particular days, some team members preferred to never come to the office. In comparison to individuals who frequently worked from the office, informants said it was harder to get to know the remote team members. Due to significant obstacles to initiating conversations when working remotely, the hybrid teams spent more time on casual interactions when co-located than before the pandemic. Those who preferred to not co-locate were thus excluded from these interactions.

3.4 Team Learning

At PubTrans, team learning was the fourth major challenge in hybrid teams. Due to the limited and mismatched co-location time, many people had difficulty asking questions and transferring knowledge. Particularly, the sharing of domain knowledge and tacit knowledge was cited by almost every informant as a significant challenge. The project’s large scope necessitated the integration of a vast number of teams, subsystems, stakeholders, and technology. Understanding the PubTrans domain therefore required an understanding of the organisation as a whole, and this knowledge was frequently tacit and undocumented. Almost all of the informants said that obtaining domain knowledge was more challenging than learning specific technologies and programming languages. While there were many internet resources for specific programming languages, finding answers to inquiries about the PubTrans domain was impossible.

Table 1. Coordination mechanisms managing organizational debt
Full size table

4 Discussion

We will now discuss our research question: “What coordination mechanisms are used to manage organizational debt challenges in large-scale agile?” In large-scale agile environments, agile practices are often used together with other organizational practices [5]. This was reflected in our findings, as the coordination mechanisms identified included both agile coordination practices, such as stand-up meetings, retrospective meetings and task boards, as well as non-agile practices like social events and gaming.

In total, we found 23 coordination mechanisms used to manage organizational debt challenges. We have categorized them into eight general coordination mechanisms and if they were affecting the aspects: Shared mental models, team coordination, team cohesion and team learning, see Table 1. Some coordination mechanisms are closely interrelated - for instance, creating shared mental models also improves the team cohesion, which in turn encourages team learning.

Shared mental models represent knowledge held in common by members that lets them understand tasks and relationship among tasks, and coordinate their actions and interactions [7]. Our findings suggested that hybrid teams should explicitly discuss their hybrid work processes, in order to create a shared mental model amongst its members. This is in line with other research [18, 21].

The coordination mechanisms co-location rules and Slack norms and etiquette, helped by explicitly discussing and describing the hybrid collaboration pattern within the teams. Discussing common co-location rules created a shared understanding of the team’s work location on a given work day. The team members could thus anticipate one another’s needs, and adjust their co-location plans accordingly. Our work builds on the findings that Slack is an essential tool for coordination in distributed teams [20]. Further, in our teams, the shared Slack etiquette created a common understanding of the availability of team members, by for example agreeing on traditions such as sending “good morning” messages when ready for work, and changing Slack icons when unavailable.

We found a set of coordination mechanisms that teams might utilize to increase team cohesion, which could improve the overall team performance [7]. These mechanisms included social activities with a common purpose, such as multiplayer games, daily quizzes, physical social events with the team and the company, and informal conversations during meetings. We found coordination mechanisms employed to encourage team learning. These mechanisms generally focused on lowering barriers to asking questions within and across teams. Frequent communication within teams are shown to improve psychological safety, which in turn lowers the barriers to asking questions and encourages knowledge sharing [4, 13].

In our study, teams worked in a hybrid setting. Unlike co-located large-scale agile, hybrid organizations must align co-location rules across teams to establish a shared mental model organization-wide. To ease tension between team autonomy and inter-team alignment, creating incentives for co-location can be beneficial. Enforcing company-wide rules, such as co-location on a set day, may threaten team autonomy in a large-scale agile organization [12]. Instead, incentives could be used to encourage co-location. For instance, PubTrans introduced the “baked goods trolley” on Thursdays, strengthening the employees’ willingness for co-location.

5 Conclusion

In conclusion, our research highlights the importance of addressing not only technical debt but also other types of organizational debt, such as process and social debt, in software development organizations, especially in the context of large-scale hybrid settings. We identified 23 coordination mechanisms that were used to manage organizational debt. By raising awareness of these non-technical forms of debt and the coordination mechanisms to address them, we aim to provide practitioners with insights to improve their software development processes and enhance overall organizational performance. Further research is encouraged to uncover additional strategies that could be beneficial in managing the complexities of large-scale, hybrid agile organizations.